Selenium is an essential trace dietary nutrient that may actively protect against damage from free radicals and reactive oxygen species produced during oxidative stress; for example, through its role as an essential component of numerous selenoproteins in critical antioxidant systems [¹,²]. Despite its central role in antioxidant defense systems and putative involvement in the development of chronic diseases, the major determinants of selenium levels in free-living humans are not well-understood. Of course selenium is consumed in the diet, but non-dietary factors, such as gastrointestinal absorption, tissue uptake, protein incorporation, or tissue metabolism or depletion, could also affect concentrations in the body. For example, older age [³], alcohol intake [⁴], and cigarette smoking [^3-6] have been associated with lower selenium levels, suggesting that non-dietary individual or environmental factors (e.g., related to oxidative stress) may alter selenium stores. However, these prior studies were generally small and did not fully adjust for dietary consumption or other potential confounding factors. Investigations through large studies have been limited by challenges in estimating selenium consumption from dietary questionnaires due to significant variation in selenium contents of similar appearing foods [³,⁷]. Additionally, measuring selenium contents of foods is limited by considerable geographic variability, and more importantly, only identifies potential selenium sources rather than what is actually consumed in the population. For example, a food that contains a high selenium level but is rarely consumed will contribute minimally to selenium exposure in the population. Advanced neutron activation analysis (NAA) measurements now allow precise measurement of selenium levels in toenails [⁸,⁹], providing an objective biomarker of long-term exposure in populations. Toenail selenium concentrations are well correlated with selenium intake as well as selenium levels in serum or whole blood [⁸,⁹]. In addition to individual behavioral determinants, geographic variation could play a strong role in selenium exposure. Selenium levels in foods vary directly with concentrations in the soil or water from which the food is derived [¹⁰,¹¹]. Conversely, in modern industrialized nations, foods are grown, processed, and transported from around the country and indeed the globe, potentially eliminating any significant residual geographic variation in exposure compared to, for example, developing nations. Unfortunately, most previous studies of selenium levels in free-living populations have been conducted in geographically relatively limited cohorts [³,⁶,⁷,¹²] limiting ability to evaluate any independent contributions of geographic variation.

To elucidate each of these issues, we investigated the independent dietary, demographic, lifestyle, and geographic determinants of selenium levels, as assessed by objective biomarker levels in toenails, in 3,902 U.S. men and women in 2 large U.S. cohorts. Our main aims were to determine [¹] the major independent dietary predictors of selenium levels [²]; whether lifestyle and demographic factors that might influence oxidative stress, such as age, smoking, adiposity, physical activity, and alcohol use, are independent predictors of selenium levels after adjusting for diet, and [³] whether significant geographic variation would remain after accounting for differences in diet, demographic, and lifestyle factors.

The average age of participants was 60 ± 9 y in men and 53 ± 6 y in women (Table 1). Approximately one-half of the study participants were overweight or obese, and 11% of men and 39% of women were current smokers. Selenium supplement use was uncommon, used by only 7% of men and 3% of women; multivitamin supplements were more common, taken by 45% of men and 42% of women. Linear spline models suggested linear relationships between toenail selenium and covariates; thus, these were evaluated as continuous variables in the multivariable models.

Among the 21 food groups evaluated, consumption of beef, breads (including cakes, muffins, and other breads), coffee, dairy products, eggs, and white rice were identified in multivariable stepwise regression models as independent determinants of selenium levels. Consumption of beef and bread products were each associated with higher levels, whereas consumption of coffee, dairy products, eggs and white rice were each associated with lower levels.

To understand whether other demographic and lifestyle factors were associated with selenium after accounting for diet, we then performed multivariable-adjusted analyses including adjustments for each of the identified dietary factors, and multivariable-adjusted results were similar in men and women, so the combined results are shown, adjusted for gender (Table 2). Even adjusting for diet as well as other lifestyle and environmental factors, both age and gender were independent determinants of selenium levels. Selenium levels were 0.62% lower for every 5 years of older age (95% CI: 0.29, 0.96), and were 6.3% lower in women vs. men (95% CI: 4.9, 7.6). This sex difference was consistent in analyses excluding current smokers and stratified by geographic region (data not shown).

Several key lifestyle factors also remained independently associated with selenium levels after adjustment for demographics and diet (Table 2). Higher BMI was associated with higher selenium levels, with 1.2% higher levels for each 5 kg/m² (95% CI: 0.3, 2.2). Compared with never smokers, current smokers had significantly lower selenium levels, with an additional dose-response relationship for the number of cigarettes smoked: selenium levels were 5% lower in those smoking 1-14 cigarettes per day, 9% lower in those smoking 15-34 cigarettes per day, and 10% lower in those smoking 35 or more cigarettes per day. Interestingly, former smokers did not have significantly different selenium levels than never smokers. Alcohol consumption was also associated with selenium levels, with 1.5% lower levels for each drink per day (95% CI: 1.1, 2.0). As expected, the use of selenium supplements was associated with 6.9% higher selenium (95% CI: 4.3, 9.5). The use of multivitamin supplements was not associated with selenium levels.

Even after adjustment for each of these demographic, lifestyle, and dietary characteristics, significant geographic variation in selenium levels was identified (Table 2). Compared to participants residing in Southern Midwestern states, those living in the West, Midwest-North, and South had significantly higher toenail selenium levels (West: 8.9% higher, 95% CI: 7.2, 10.6; Northern Midwest: 7.4% higher, 95% CI: 5.5, 9.4; and South: 2.8% higher, 95% CI: 0.9, 4.7). There was little evidence that these relationships varied by sex, smoking, or supplement use. Results remained identical after additional adjustment for total energy intake.

Adjusted mean selenium levels in men in individual U.S. states are shown in Fig. 1. Overall, men in the northwestern states generally had higher levels of selenium than those in the southeastern states. The highest levels of selenium were found in Wyoming, South and North Dakota, Nebraska, and Washington state; the adjusted means of toenail selenium in these states were 1.00, 0.97, 0.96, 0.94, and 0.94 µg/g, respectively. States with the lowest selenium levels were North Carolina, Kentucky, Pennsylvania, Nevada, and South Carolina, in which adjusted selenium means were only 0.77, 0.77, 0.77, 0.76, and 0.75 µg/g, respectively. Average levels were lower but geographic differences similar in women in 11 major U.S. states (data not shown). There were no significant effect modifications between smoking, sex, and supplement use.

In this well-characterized population of U.S. men and women, multiple independent determinants of toenail selenium levels were identified. It would be expected that selenium levels in the body should be affected by dietary factors and supplement use, and both were identified here as independent determinants. However, although the magnitude of these observed relationships is likely to be underestimated due to some misclassification in estimation of diet, even so these dietary factors explained only a modest relative proportion of overall variation in selenium levels. Notably, several other demographic and lifestyle factors were independent determinants of selenium levels, including smoking, alcohol intake, age, and sex. Furthermore, geographic residence remained a strong independent predictor, even after adjustment for types of foods consumed and each of the other independent determinants of toenail selenium.

Beef and bread products are recognized as good sources of selenium [²⁶], and bioavailability of selenium from these foods is relatively high [²⁷,²⁸]. However, consumption of white rice was also negatively correlated with selenium levels in our study. This could be explained by findings that consumption of high glycemic index (GI) foods leads to higher oxidative stress [²⁹,³⁰]. Selenoproteins are important components of several critical antioxidant systems, such as glutathione peroxidase, that actively protect against damage from free radicals and reactive oxygen species [³¹,³²]. Thus, it is possible that consuming foods that induce oxidative stress, such as high GI foods, could deplete selenium levels. It is also possible that foods such as white rice, which are typically refined and depleted of phytochemicals, could be replacing other higher-selenium foods in the diet. Interestingly, intakes of coffee, dairy products, and eggs were each inversely associated with selenium levels. No prior study has evaluated intakes of these foods in relation to selenium levels. Our findings support the need for further study of potential effects of these foods on selenium levels.

We observed a dose-response relationship between the number of cigarettes currently smoked and toenail selenium levels, consistent with findings from previous studies [^3-5], even after multivariable adjustment for dietary habits, demographics, and other lifestyle factors. However, why cigarette smokers have lower levels of selenium has not been established. Swanson et al. [⁶] and van den Brandt et al. [⁵] reported that compared to nonsmokers, current smokers have lower consumption of selenium as estimated from dietary questionnaires, suggesting that lower selenium levels in smokers could at least partly reflect dietary differences. However, we identified current smoking as a significant predictor of lower selenium levels even after adjustment for other dietary and non-dietary determinants. Although residual confounding from unmeasured dietary factors and measurement error in the observed dietary factors cannot be excluded, the magnitude and observed dose-response support biological plausibility of a causal relationship. Cigarette smoking is known to increase oxidative stress [^33-35], with a positive dose-response relationship per number of cigarettes smoked [³⁶], and biomarkers of oxidative stress are reduced substantially within 2 weeks of smoking cessation [³⁴]. These results support our findings of lower selenium levels in current but not former smokers, and the lowest levels among those who smoked the most cigarettes per day. Our findings indicate that factors which promote oxidative stress, such as current smoking, may directly affect tissue selenium levels, for example by increasing selenium utilization and causing depletion of tissue selenium stores. This concept of selenium depletion is further supported by the relationships we found between some other non-dietary characteristics and selenium levels. For example, older age was associated with lower toenail selenium levels, consistent with previous reports [³,³⁷], which may be explained by increased oxidative damage as a function of aging [³⁸,³⁹]. In addition, we observed alcohol-related reductions in selenium levels, a finding supported by reports that alcohol consumption induces oxidative stress [⁴⁰,⁴¹]. Thus, our findings together support a new hypothesis, in which factors that promote oxidative stress may deplete selenium levels in the body.

Prior studies of selenium levels have found inconsistent sex differences [^4-6,^42-44]. The sex differences we observed should be interpreted conservatively since these were two separately recruited cohorts. Interestingly, prior findings have also been inconsistent in terms of whether oxidative stress levels are higher in men or women [⁴⁵]. The reasons for these discrepancies in relationships between sex and both selenium and oxidative stress status are unknown, and further investigation of these issues is warranted.

Although the extensive transport system of food products in the United States is thought to reduce variability in the selenium contents of foods in different regions, significant geographic heterogeneity in selenium levels was observed, even after accounting for dietary habits, demographics, and other lifestyle factors. This could be due to significant variation in the selenium contents of foods in different U.S. regions, or to residual confounding from unmeasured differences in other dietary habits or lifestyle characteristics between regions. Whereas our findings cannot definitively distinguish between these two possibilities, our evaluation and adjustment for multiple dietary habits, demographic characteristics, and lifestyle factors suggests that the former explanation may be more plausible. Previous studies have shown inconsistent geographic distributions of selenium in crops and human levels in the U.S. [³,⁴²,⁴⁶,⁴⁷], and most prior studies in humans [³,⁴²] have not reported the independent effect of geographic variation adjusted for other potential selenium determinants, including diet and lifestyle factors. Our study provides unique geographic information about the distribution of selenium levels, adjusted for other major determinants, in two large U.S. cohorts.

This analysis has several strengths. Rather than measuring selenium content in foods and estimating exposure, we assessed actual tissue biomarkers of selenium levels in 3,902 free-living men and women from across the U.S., providing a better estimate of actual exposure to selenium in the population. Selenium levels were evaluated in toenails using well-established methods, providing a valid and reproducible biomarker of long-term levels. The cohort size and standardized assessment of a wide variety of participant characteristics allowed multivariable-adjustment for several demographic, lifestyle, and dietary factors, including multiple potential dietary sources of selenium, minimizing residual confounding. We concurrently evaluated multiple potential determinants, providing the best evidence to-date of how each of these demographic, lifestyle, and dietary factor may independently influence long-term selenium levels.

Limitations should also be considered. The participants were Americans of higher socioeconomic status and mostly white, and our findings may not be generalizable to other socioeconomic groups or nations. On the other hand, there is little reason to expect that the biological effects on selenium levels of some factors, such as smoking, might vary substantially in other socioeconomic groups, races, or countries. These cohorts do not represent random samples of the U.S. population, and thus the geographic findings should be viewed qualitatively rather than as definitive for the U.S. population. Although extensive covariate information was collected using validated instruments, some residual confounding due to unmeasured or imperfectly measured factors may remain.

In summary, we identified several independent dietary, and interestingly, non-dietary predictors of selenium levels in the U.S. population. For many of the non-dietary factors, the findings were consistent with expected effects of oxidative stress, suggesting that tissue selenium, and potentially selenium-related anti-oxidant defenses, may depend on the balance between dietary consumption vs. depletion by oxidant stressors. Geographic variation was also an independent determinant, suggesting that even in the U.S. substantial regional differences in the selenium contents of foods may be present. Finally, although we identified a number of independent determinants of selenium-providing critical information for future studies examining selenium levels and chronic disease - a large proportion of the total variance was left unexplained. Future studies, including the examination of genetic factors, are needed to better understand the biological pathways through which environmental and genetic determinants affect selenium concentrations in the human population.

This study was supported by NIH grant R01 ES 014433 (NIEHS and NHLBI) and the Harvard School of Public Health Genes and Environment Initiative (GENI), with additional support from NIH grants CA55075 and HL35464 and the Yeungnam University research grant in 2011.